Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Mitigating the Uncanny Valley Effect in Hyper-Realistic Robots: A Student-Centered Study on LLM-Driven Conversations

Hangyeol Kang, Thiago Freitas dos Santos, Maher Ben Moussa, Nadia Magnenat-Thalmann

TL;DR

The paper addresses the uncanny valley barrier facing hyper-realistic social robots and tests whether LLM-driven dialogue can reduce eeriness and improve engagement. Using Nadine, with a simplified SoR-ReAct architecture and LangGraph modules (router, retriever, web search, generator) and an LLM (gpt-4o-mini) plus a Chroma vectorstore, the study delivers context-aware conversations. In a one-on-one study with 68 university participants, pre- and post-interaction surveys measured creepiness, pleasantness, approachability, naturalness, and interestingness; regression analyses identified naturalness and interestingness as key predictors of willingness to continue, while human-likeness was not. The findings suggest design guidance that prioritizes fluid, engaging dialogue over purely human-like mimicry, highlighting the practical potential of LLMs to bridge the uncanny valley in real-world social robotics.

Abstract

The uncanny valley effect poses a significant challenge in the development and acceptance of hyper-realistic social robots. This study investigates whether advanced conversational capabilities powered by large language models (LLMs) can mitigate this effect in highly anthropomorphic robots. We conducted a user study with 80 participants interacting with Nadine, a hyper-realistic humanoid robot equipped with LLM-driven communication skills. Through pre- and post-interaction surveys, we assessed changes in perceptions of uncanniness, conversational quality, and overall user experience. Our findings reveal that LLM-enhanced interactions significantly reduce feelings of eeriness while fostering more natural and engaging conversations. Additionally, we identify key factors influencing user acceptance, including conversational naturalness, human-likeness, and interestingness. Based on these insights, we propose design recommendations to enhance the appeal and acceptability of hyper-realistic robots in social contexts. This research contributes to the growing field of human-robot interaction by offering empirical evidence on the potential of LLMs to bridge the uncanny valley, with implications for the future development of social robots.

Mitigating the Uncanny Valley Effect in Hyper-Realistic Robots: A Student-Centered Study on LLM-Driven Conversations

TL;DR

The paper addresses the uncanny valley barrier facing hyper-realistic social robots and tests whether LLM-driven dialogue can reduce eeriness and improve engagement. Using Nadine, with a simplified SoR-ReAct architecture and LangGraph modules (router, retriever, web search, generator) and an LLM (gpt-4o-mini) plus a Chroma vectorstore, the study delivers context-aware conversations. In a one-on-one study with 68 university participants, pre- and post-interaction surveys measured creepiness, pleasantness, approachability, naturalness, and interestingness; regression analyses identified naturalness and interestingness as key predictors of willingness to continue, while human-likeness was not. The findings suggest design guidance that prioritizes fluid, engaging dialogue over purely human-like mimicry, highlighting the practical potential of LLMs to bridge the uncanny valley in real-world social robotics.

Abstract

The uncanny valley effect poses a significant challenge in the development and acceptance of hyper-realistic social robots. This study investigates whether advanced conversational capabilities powered by large language models (LLMs) can mitigate this effect in highly anthropomorphic robots. We conducted a user study with 80 participants interacting with Nadine, a hyper-realistic humanoid robot equipped with LLM-driven communication skills. Through pre- and post-interaction surveys, we assessed changes in perceptions of uncanniness, conversational quality, and overall user experience. Our findings reveal that LLM-enhanced interactions significantly reduce feelings of eeriness while fostering more natural and engaging conversations. Additionally, we identify key factors influencing user acceptance, including conversational naturalness, human-likeness, and interestingness. Based on these insights, we propose design recommendations to enhance the appeal and acceptability of hyper-realistic robots in social contexts. This research contributes to the growing field of human-robot interaction by offering empirical evidence on the potential of LLMs to bridge the uncanny valley, with implications for the future development of social robots.

Paper Structure

This paper contains 13 sections, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Uncanny Valley in Humanoid Robots
  4. Large Language Models and Human-Robot Interaction
  5. Nadine Social Robot
  6. Nadine Platform
  7. Method
  8. Participants
  9. Experimental setup and procedure
  10. Analysis
  11. Results
  12. Discussion
  13. Conclusion

Figures (3)

  • Figure 1: Nadine system architecture.
  • Figure 2: Boxplots depicting changes in students' ratings for pleasantness, creepiness, and approachability before (pre) and after (post) interacting with the robot. Each box represents the interquartile range (IQR), with the median shown as a horizontal line and individual data points overlaid.
  • Figure 3: Regression coefficients with standard errors for the predictors of naturalness, human-likeness, and interestingness, along with the constant term.